Benjamin C. Wagner

High dimensional classification of structural MRI Alzheimer's disease data based on large scale regularization.

In this work we use a large scale regularization approach based on penalized logistic regression to automatically classify structural MRI images (sMRI) according to cognitive status. Its performance is illustrated using sMRI data from the Alzheimer Disease Neuroimaging Initiative (ADNI) clinical database. We downloaded sMRI data from 98 subjects (49 cognitive normal and 49 patients) matched by age and sex from the ADNI website. Images were segmented and normalized using SPM8 and ANTS software packages. Classification was performed using GLMNET library implementation of penalized logistic regression based on coordinate-wise descent optimization techniques. To avoid optimistic estimates classification accuracy, sensitivity, and specificity were determined based on a combination of three-way split of the data with nested 10-fold cross-validations. One of the main features of this approach is that classification is performed based on large scale regularization. The methodology presented here was highly accurate, sensitive, and specific when automatically classifying sMRI images of cognitive normal subjects and Alzheimer disease (AD) patients. Higher levels of accuracy, sensitivity, and specificity were achieved for gray matter (GM) volume maps (85.7, 82.9, and 90%, respectively) compared to white matter volume maps (81.1, 80.6, and 82.5%, respectively). We found that GM and white matter tissues carry useful information for discriminating patients from cognitive normal subjects using sMRI brain data. Although we have demonstrated the efficacy of this voxel-wise classification method in discriminating cognitive normal subjects from AD patients, in principle it could be applied to any clinical population.

Cerebral Structural Changes in Diabetic Kidney Disease: African American-Diabetes Heart Study MIND

OBJECTIVE Albuminuria and reduced kidney function are associated with cognitive impairment. Relationships between nephropathy and cerebral structural changes remain poorly defined, particularly in African Americans (AAs), a population at higher risk for both cognitive impairment and diabetes than European Americans. We examined the relationship between urine albumin:creatinine ratio (UACR), estimated glomerular filtration rate (eGFR), and cerebral MRI volumes in 263 AAs with type 2 diabetes. RESEARCH DESIGN AND METHODS Cross-sectional associations between renal parameters and white matter (WM), gray matter (GM), hippocampal, and WM lesion (WML) volumes were assessed using generalized linear models adjusted for age, education, sex, BMI, hemoglobin A1c (HbA1c) level, and hypertension. RESULTS Participants had a mean (SD) age of 60.2 years (9.7 years), and 62.7% were female. Mean diabetes duration was 14.3 years (8.9 years), HbA1c level was 8.2% (2.2%; 66 mmol/mol), eGFR was 86.0 mL/min/1.73 m2 (23.2 mL/min/1.73 m2), and UACR was 155.8 mg/g (542.1 mg/g; median 8.1 mg/g). Those with chronic kidney disease (CKD) (eGFR <60 mL/min/1.73 m2 or UACR >30 mg/g) had smaller GM and higher WML volumes. Higher UACR was significantly associated with higher WML volume and greater atrophy (larger cerebrospinal fluid volumes), and smaller GM and hippocampal WM volumes. A higher eGFR was associated with larger hippocampal WM volumes. Consistent with higher WML volumes, participants with CKD had significantly poorer processing speed and working memory. These findings were independent of glycemic control. CONCLUSIONS We found albuminuria to be a better marker of cerebral structural changes than eGFR in AAs with type 2 diabetes. Relationships between albuminuria and brain pathology may contribute to poorer cognitive performance in patients with mild CKD.

Combining Graph and Machine Learning Methods to Analyze Differences in Functional Connectivity Across Sex

In this work we combine machine learning methods and graph theoretical analysis to investigate gender associated differences in resting state brain network connectivity. The set of all correlations computed from the fMRI resting state data is used as input features for classification. Two ensemble learning methods are used to perform the detection of the set of discriminative edges between groups (males vs. females) of brain networks: 1) Random Forest and 2) an ensemble method based on least angle shrinkage and selection operator (lasso) regressors. Permutation testing is used not only to assess significance of classification accuracy but also to evaluate significance of feature selection. Finally, these methods are applied to data downloaded from the Connectome Project website. Our results suggest that gender differences in brain function may be related to sexually dimorphic regional connectivity between specific critical nodes via gender-discriminative edges.

Application of machine learning methods to describe the effects of conjugated equine estrogens therapy on region-specific brain volumes

Use of conjugated equine estrogens (CEE) has been linked to smaller regional brain volumes in women aged ≥65 years; however, it is unknown whether this results in a broad-based characteristic pattern of effects. Structural magnetic resonance imaging was used to assess regional volumes of normal tissue and ischemic lesions among 513 women who had been enrolled in a randomized clinical trial of CEE therapy for an average of 6.6 years, beginning at ages 65-80 years. A multivariate pattern analysis, based on a machine learning technique that combined Random Forest and logistic regression with L(1) penalty, was applied to identify patterns among regional volumes associated with therapy and whether patterns discriminate between treatment groups. The multivariate pattern analysis detected smaller regional volumes of normal tissue within the limbic and temporal lobes among women that had been assigned to CEE therapy. Mean decrements ranged as high as 7% in the left entorhinal cortex and 5% in the left perirhinal cortex, which exceeded the effect sizes reported previously in frontal lobe and hippocampus. Overall accuracy of classification based on these patterns, however, was projected to be only 54.5%. Prescription of CEE therapy for an average of 6.6 years is associated with lower regional brain volumes, but it does not induce a characteristic spatial pattern of changes in brain volumes of sufficient magnitude to discriminate users and nonusers.

Subclinical Atherosclerosis Is Inversely Associated With Gray Matter Volume in African Americans With Type 2 Diabetes

OBJECTIVE Relative to European Americans, African Americans manifest lower levels of computed tomography–based calcified atherosclerotic plaque (CP), a measure of subclinical cardiovascular disease (CVD). Potential relationships between CP and cerebral structure are poorly defined in the African American population. We assessed associations among glycemic control, inflammation, and CP with cerebral structure on MRI and with cognitive performance in 268 high-risk African Americans with type 2 diabetes. RESEARCH DESIGN AND METHODS Associations among hemoglobin A1c (HbA1c), C-reactive protein (CRP), and CP in coronary arteries, carotid arteries, and the aorta with MRI volumetric analysis (white matter volume, gray matter volume [GMV], cerebrospinal fluid volume, and white matter lesion volume) were assessed using generalized linear models adjusted for age, sex, African ancestry proportion, smoking, BMI, use of statins, HbA1c, hypertension, and prior CVD. RESULTS Participants were 63.4% female with mean (SD) age of 59.8 years (9.2), diabetes duration of 14.5 years (7.6), HbA1c of 7.95% (1.9), estimated glomerular filtration rate of 86.6 mL/min/1.73 m2 (24.6), and coronary artery CP mass score of 215 mg (502). In fully adjusted models, GMV was inversely associated with coronary artery CP (parameter estimate [β] −0.47 [SE 0.15], P = 0.002; carotid artery CP (β −1.92 [SE 0.62], P = 0.002; and aorta CP [β −0.10 [SE 0.03] P = 0.002), whereas HbA1c and CRP did not associate with cerebral volumes. Coronary artery CP also associated with poorer global cognitive function on the Montreal Cognitive Assessment. CONCLUSIONS Subclinical atherosclerosis was associated with smaller GMV and poorer cognitive performance in African Americans with diabetes. Cardioprotective strategies could preserve GMV and cognitive function in high-risk African Americans with diabetes.

Associations between anxiety and depression symptoms and cognitive testing and neuroimaging in type 2 diabetes

AIMS Anxiety, depression, accelerated cognitive decline, and increased risk of dementia are observed in individuals with type 2 diabetes. Anxiety and depression may contribute to lower performance on cognitive tests and differences in neuroimaging observed in individuals with type 2 diabetes. METHODS These relationships were assessed in 655 European Americans with type 2 diabetes from 504 Diabetes Heart Study families. Participants completed cognitive testing, brain magnetic resonance imaging, the Brief Symptom Inventory Anxiety subscale, and the Center for Epidemiologic Studies Depression-10. RESULTS In analyses adjusted for age, sex, educational attainment, and use of psychotropic medications, individuals with comorbid anxiety and depression symptoms had lower performance on all cognitive testing measures assessed (p≤0.005). Those with both anxiety and depression also had increased white matter lesion volume (p=0.015), decreased gray matter cerebral blood flow (p=4.43×10(-6)), decreased gray matter volume (p=0.002), increased white and gray matter mean diffusivity (p≤0.001), and decreased white matter fractional anisotropy (p=7.79×10(-4)). These associations were somewhat attenuated upon further adjustment for health status related covariates. CONCLUSIONS Comorbid anxiety and depression symptoms were associated with cognitive performance and brain structure in a European American cohort with type 2 diabetes.

Associations of Early Kidney Disease With Brain Magnetic Resonance Imaging and Cognitive Function in African Americans With Type 2 Diabetes Mellitus

BACKGROUND Relationships between early kidney disease, neurocognitive function, and brain anatomy are poorly defined in African Americans with type 2 diabetes mellitus (T2DM). STUDY DESIGN Cross-sectional associations were assessed between cerebral anatomy and cognitive performance with estimated glomerular filtration rate (eGFR) and urine albumin-creatinine ratio (UACR) in African Americans with T2DM. SETTING & PARTICIPANTS African Americans with cognitive testing and cerebral magnetic resonance imaging (MRI) in the African American-Diabetes Heart Study Memory in Diabetes (AA-DHS MIND; n=512; 480 with MRI) and Action to Control Cardiovascular Risk in Diabetes (ACCORD) MIND (n=484; 104 with MRI) studies. PREDICTORS eGFR (CKD-EPI creatinine equation), spot UACR. MEASUREMENTS MRI-based cerebral white matter volume (WMV), gray matter volume (GMV), and white matter lesion volume; cognitive performance (Mini-Mental State Examination, Digit Symbol Coding, Stroop Test, and Rey Auditory Verbal Learning Test). Multivariable models adjusted for age, sex, body mass index, scanner, intracranial volume, education, diabetes duration, hemoglobin A1c concentration, low-density lipoprotein cholesterol concentration, smoking, hypertension, and cardiovascular disease were used to test for associations between kidney phenotypes and the brain in each study; a meta-analysis was performed. RESULTS Mean participant age was 60.1±7.9 (SD) years; diabetes duration, 12.1±7.7 years; hemoglobin A1c concentration, 8.3%±1.7%; eGFR, 88.7±21.6mL/min/1.73m2; and UACR, 119.2±336.4mg/g. In the fully adjusted meta-analysis, higher GMV associated with lower UACR (P<0.05), with a trend toward association with higher eGFR. Higher white matter lesion volume was associated with higher UACR (P<0.05) and lower eGFR (P<0.001). WMV was not associated with either kidney parameter. Higher UACR was associated with lower Digit Symbol Coding performance (P<0.001) and a trend toward association with higher Stroop interference; eGFR was not associated with cognitive tests. LIMITATIONS Cross-sectional; single UACR measurement. CONCLUSIONS In African Americans with T2DM, mildly high UACR and mildly low eGFR were associated with smaller GMV and increased white matter lesion volume. UACR was associated with poorer processing speed and working memory.

Using Convolutional Neural Networks to Automatically Detect Eye-Blink Artifacts in Magnetoencephalography Without Resorting to Electrooculography

Magnetoencephelography (MEG) is a functional neuroimaging tool that records the magnetic fields induced by neuronal activity; however, signal from muscle activity often corrupts the data. Eye-blinks are one of the most common types of muscle artifact. They can be recorded by affixing eye proximal electrodes, as in electrooculography (EOG), however this complicates patient preparation and decreases comfort. Moreover, it can induce further muscular artifacts from facial twitching. We propose an EOG free, data driven approach. We begin with Independent Component Analysis (ICA), a well-known preprocessing approach that factors observed signal into statistically independent components. When applied to MEG, ICA can help separate neuronal components from non-neuronal ones, however, the components are randomly ordered. Thus, we develop a method to assign one of two labels, non-eye-blink or eye-blink, to each component. Our contributions are two-fold. First, we develop a 10-layer Convolutional Neural Network (CNN), which directly labels eye-blink artifacts. Second, we visualize the learned spatial features using attention mapping, to reveal what it has learned and bolster confidence in the methods ability to generalize to unseen data. We acquired 8-min, eyes open, resting state MEG from 44 subjects. We trained our method on the spatial maps from ICA of 14 subjects selected randomly with expertly labeled ground truth. We then tested on the remaining 30 subjects. Our approach achieves a test classification accuracy of 99.67%, sensitivity: 97.62%, specificity: 99.77%, and ROC AUC: 98.69%. We also show the learned spatial features correspond to those human experts typically use which corroborates our models validity. This work (1) facilitates creation of fully automated processing pipelines in MEG that need to remove motion artifacts related to eye blinks, and (2) potentially obviates the use of additional EOG electrodes for the recording of eye-blinks in MEG studies.

Changes in resting state MRI networks from a single season of football distinguishes controls, low, and high head impact exposure

Sub-concussive asymptomatic head impacts during contact sports may develop potential neurological changes and may have accumulative effect through repetitive occurrences in contact sports like American football. The effects of sub-concussive head impacts on the functional connectivity of the brain are still unclear with no conclusive results yet presented. Although various studies have been performed on the topic, they have yielded mixed results with some concluding that sub concussive head impacts do not have any effect on functional connectivity, while others concluding that there are acute to chronic effects. The purpose of this study is to determine whether there is an effect on the functional connectivity of the brain from repetitive sub concussive head impacts. First, we applied a model free group ICA based intrinsic network selection to consider the relationship between all voxels while avoiding an arbitrary choice of seed selection. Second, unlike most other studies, we have utilized the default mode network along with other extracted intrinsic networks for classification. Third, we systematically tested multiple supervised machine learning classification algorithms to predict whether a player was a non-contact sports youth player, a contact sports player with low levels of cumulative biomechanical force impacts, or one with high levels of exposure. The 10-fold cross validation results show robust classification between the groups with accuracy up to 78% establishing the potential of data driven approaches coupled with machine learning to study connectivity changes in youth football players. This work adds to the growing body of evidence that there are detectable changes in brain signature from playing a single season of contact sports.

APOE Genotypes Associate With Cognitive Performance but Not Cerebral Structure: Diabetes Heart Study MIND.

OBJECTIVE Dementia is a debilitating illness with a disproportionate burden in patients with type 2 diabetes (T2D). Among the contributors, genetic variation at the apolipoprotein E locus (APOE) is posited to convey a strong effect. This study compared and contrasted the association of APOE with cognitive performance and cerebral structure in the setting of T2D. RESEARCH DESIGN AND METHODS European Americans from the Diabetes Heart Study (DHS) MIND (n = 754) and African Americans from the African American (AA)-DHS MIND (n = 517) were examined. The cognitive battery assessed executive function, memory, and global cognition, and brain MRI was performed. RESULTS In European Americans and African Americans, the APOE E4 risk haplotype group was associated with poorer performance on the modified Mini-Mental Status Examination (P < 0.017), a measure of global cognition. In contrast to the literature, the APOE E2 haplotype group, which was overrepresented in these participants with T2D, was associated with poorer Rey Auditory Verbal Learning Test performance (P < 0.032). Nominal associations between APOE haplotype groups and MRI-determined cerebral structure were observed. CONCLUSIONS Compared with APOE E3 carriers, E2 and E4 carriers performed worse in the cognitive domains of memory and global cognition. Identification of genetic contributors remains critical to understanding new pathways to prevent and treat dementia in the setting of T2D.